An Introduction to Unification- Based Approaches to Grammar The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters Citation Shieber, Stuart M. 2003. An introduction to unification-based approaches to grammar. Brookline, Massachusetts: Microtome Publishing. Reissue of Shieber, Stuart M. 1986. An introduction to unification-based approaches to grammar. Stanford, California: CSLI Publications. Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:11576719 Terms of Use This article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- use#LAA An Introduction to Unification-Based Approaches to Grammar i An Introduction to Unification-Based Approaches to Grammar Stuart M. Shieber Microtome Publishing Brookline, Massachusetts c Stuart M. Shieber 1986, 1988, 2003 This work by Stuart M. Shieber is licensed under a Creative Commons Attribution 3.0 Unported License. The full text of the license is available at http://creativecommons.org/licenses/by/3.0/deed.en_US. PREFACE v Preface These notes were originally developed as a supplement to a tutorial on unification-based approaches to grammar presented at the 23rd Annual Meet- ing of the Association for Computational Linguistics on 7 July, 1985 at the University of Chicago. My intention was to present a set of formalisms and theories from a single organizing perspective, that of their reliance on notions of identity of complex feature structures and on unification in the underlying algebras. Of course, as discussed in the first chapter, such an expositional method precludes cover- ing the formalisms from the perspective put forward by particular researchers in the field. Fortunately, I can now recommend an introduction to the two main linguistic theories discussed in these notes which does present their own view in an introductory way yet with surprisingly broad coverage; Peter Sells’ Lectures on Contemporary Syntactic Theories, published in this same CSLI Lecture Note Series, serves this purpose admirably. Because of the brevity of the present work, and its intended (though not completely realized) nonparochial nature, I have purposefully left out any lin- guistic analyses from a unification-based perspective of any but the most cur- sory sort. My reasoning here was that the vast linguistics literature in the area provides ample examples of such analyses. (For interested readers, Chapter 5 contains references to some of this research.) A future edition, however, may be extended with examples of analyses of control, long-distance dependencies and various of the other canonical linguistic phenomena described using the tools of unification-based formalisms. Much of my thinking on these issues has been influenced by researchers at the Center for the Study of Language and Information at Stanford University, and especially those in the Foundations of Grammar project. CSLI is a unique environment for research, and has done more to shape this work and the opin- ions presented herein than any other single force. I am also indebted to Martin Kay, Fernando Pereira, Ivan Sag, Susan Stucky, Hans Uszkoreit, Thomas Wa- sow and Annie Zaenen for their comments on earlier drafts of these notes. However, the views and conclusions expressed here are not necessarily shared by them, nor are they accountable for any errors remaining. The exposition in these notes, such as it is, was considerably improved, as usual, by my editor at SRI, Savel Kliachko. Any unfortunate idiosyncrasies in the text are undoubtedly lapses of the author; Savel never errs. I also extend vi heartfelt thanks to the CSLI editor, Dikran Karagueuzian, for his patience and support of this work. Finally, the preparation of these notes was considerably simplified by the efforts of Emma Pease in the formatting and indexing of the final version. The preparation of the published version of these notes was made possible by a gift from the System Development Foundation. Menlo Park, California 13 February, 1986 This second printing corrects some typographical and bibliographical errors pointed out by readers of the first printing. Special thanks go to Geoffrey K. Pullum for his assiduous comments on the citations. Menlo Park, California 2 February, 1988 This reissue of “An Introduction to Unification-Based Approaches to Grammar Formalisms” is only lightly edited from the previous 1988 printing. By now, the content is considered quite elementary and the presentation is grossly out of date, but it may have some remaining utility by virtue of the former property at least. More complete and modern presentations of the formal basis for the type of formalisms described in this book can be found in the books by Carpenter (1992) and Johnson (1988). From a linguistic perspective, a good starting point is the textbook by Sag and Wasow (1999). Cambridge, Massachusetts 20 August, 2003 Contents Preface v Chapter1. IntroductionandHistory 1 Chapter2. TheUnderlyingFramework 3 2.1. TheRoleofGrammarFormalisms 3 2.2. SomeParticularDesignChoices 4 2.3. CoverageoftheFramework 5 Chapter 3. The Simplest Unification-Based Formalism 9 3.1. OverviewoftheFormalism 9 3.2. TheInformationalDomain 10 3.3. Combinatory Rules 17 3.4. SomePATR-IIGrammars 20 Chapter4. ExtendedFormalisms 31 4.1. Introduction 31 4.2. Two Classes of Formalisms 31 4.3. FunctionalUnificationGrammar 32 4.4. Definite-ClauseGrammars 37 4.5. Lexical-FunctionalGrammar 39 4.6. GeneralizedPhraseStructureGrammar 42 4.7. Head Grammar and Head-Driven Phrase Structure Grammar 44 4.8. Lexical Organization 45 4.9. OtherExtensionsofFormalisms 53 Chapter 5. Conclusions 55 5.1. EmergentIssuesConcerningFormalisms 55 5.2. A Summary 57 AppendixA. TheSamplePATR-IIGrammars 59 vii viii CONTENTS A.1. Sample Grammar One 59 A.2. Sample Grammar Two 62 A.3. Sample Grammar Three 67 A.4. Sample Grammar Four 73 Appendix B. The Literature 79 B.1. General Papers 79 B.2. BackgroundandOverviewsoftheFormalisms 79 B.3. HandlingSpecificLinguisticPhenomena 80 B.4. Related Formalisms and Languages from Computer Science 80 B.5. RelatedImplementationTechniques 81 B.6. ImplementationsofUnification-BasedFormalisms 82 Bibliography 83 CHAPTER 1 Introduction and History These notes discuss a particular approach to encoding linguistic informa- tion, both syntactic and semantic, which has been called “unification-based” or “complex-feature-based.” The term “unification-based grammar formalism” covers quite a variety of formal devices, which have been developed along historically different paths. The fact that they are regarded as a group is due perhaps partly to accidental and historical circumstances, partly to a shared underlying view of grammar, but most importantly to their reliance in some way on one particular device—unification. Historically, these grammar formalisms are the result of separate research in computational linguistics, formal linguistics, and natural-language process- ing; related techniques can be found in theorem proving, knowledge repre- sentation research, and theory of data types. Several independently initiated strains of research have converged on the idea of unification to control the flow of information. Beginning with the augmented-transition-network (ATN) concept (like so much of the research in modern computational linguistics) and inspired by Bresnan’s work on lexically oriented nontransformational linguistics, the lexical-functional grammar (LFG) framework of Bresnan and Kaplan was evolved. Simultaneously, Kay devised the functional grammar (later unifi- cation grammar, now functional unification grammar [FUG]) formalism. Independently, Colmerauer had produced the Q-system and metamorpho- sis grammar formalisms as tools for natural-language processing. The logic- programming community, specifically Pereira and Warren, created definite- clause grammars (DCG) on the basis of Colmerauer’s earlier work on these formalisms and on the programming language Prolog. Independent work in logic programminghas used DCG as the foundationof many unification-based formalisms, such as extraposition, slot, and gapping grammars. 1 2 1.INTRODUCTIONANDHISTORY Another strain of parallel research grew out of the work on nontransfor- mational linguistic analyses, from which Gazdar developed generalized phrase structure grammar (GPSG). In its later formalization by Gazdar and Pullum, GPSG imported a unification relation. Pollard defined head grammars in his dissertation, basing them on GPSG. Implementation of GPSG at Hewlett- Packard led Pollard and his colleagues to design their head-driven phrase- structure grammar (HPSG) as a successor to GPSG and head grammars. Most recently, influenced by early papers on GPSG, Rosenschein and the author devised PATR as a successor to the DIAGRAM grammar formalism at SRI International. Although PATR did not use unification, it developed (under the influence of FUG, later work in GPSG, and DCG) into PATR-II, perhaps the simplest of the unification-based formalisms. One thing has become clear through this morass of historical fortuities: unification is a powerful tool for grammar formalisms. All these formalisms make crucial use of unification in one manner or another, supplementing it, for various linguistic or computational reasons, with other mechanisms. A Caveat. Unification formalisms, as we have seen, can be traced to a diversity of sources. Consequently, proponents of each formalism have